the financial implications of corporate fraudconference/conference2012/proceedings...this paper...

i

The Financial Implications of Corporate Fraud

Chen Lin Chinese University of Hong Kong [email protected]

Frank M. Song

University of Hong Kong [email protected]

Zengyuan Sun

University of Hong Kong [email protected]

This version: Nov 1, 2012 First version: Jun 1, 2011

Abstract This paper explores the financial implications of corporate fraud by examining the impact of corporate fraud on fraudulent firms’ external financing cost and corporate cash holdings. Using a sample consist of 184 fraudulent firms that experience material litigation in securities class action, we find that firms’ cost of debt significantly increases associated with corporate fraud incident. In line with the costly external financing evidence, fraudulent firms accumulate more cash to keep liquidity and avoid underinvestment issues. Consistent with the precautionary motive argument, the value of cash increases after corporate fraud. In addition, corporate fraud contributes to financial constrains in the sense that fraudulent firms display a positive cash flow sensitivity of cash after corporate fraud. Our result indicates that corporate fraud can have a real impact on corporate outcomes by affecting the external financing cost and internal cash holdings. Keywords: Corporate fraud, Cash, Costly external finance We thank Gurdip S. Bakshi, Sreedhar T. Bharath, Xiaohui Gao, Po-Hsuan Hsu, Grace Hu, Scott Lee, Si Li, Tse-Chun Lin, Lin Peng, Dragon Tang, K.M. Wan, Chenggang Xu, Xiaoyun Yu, Xianming Zhou, and all participants of FMA Annual Meeting (2012), Financing Public and Private Firms: Fraud, Ethics, and Regulation Symposium (2012), China International Conference in Finance (2012), Asian Law and Economics Association Annual Conference (2011), seminar at University of Hong Kong (2011) for helpful comments and suggestions. We thank Peng Yun, Ma Tao, and Deng Kaizhong for data collection. Zengyuan Sun gratefully acknowledge the data support from Cornerstone Research and Stanford Law School. All the views expressed in the paper are views of the authors and do not represent in any way the views of Cornerstone Research or Stanford Law School. Zengyuan Sun also thanks Kamal R. Hubbard and Ellen Ryan for kind response and help in data collection from Stanford Securities Class Action Clearinghouse (SCAC) and Cornerstone Research.

1

1 Introduction

Corporate fraud is one of the most problematic issues for companies around the world.

Based on PricewaterhouseCooper’s Global Economic Survey 2009, of the 3037 corporate

respondents in 54 countries, 30% reported having experienced at least one incident of fraud

over the last 12 months before the survey date. In a fraud survey 2008 conducted by KPMG,

survey respondents reported 222,577 incidents of corporate fraud in Australia and New

Zealand during the survey period from February 2006 to January 2008. Corporate fraud is

also becoming prevalent in US. In the early 2000’s, triggered by Enron’s bankruptcy after

revealing massive accounting fraud, there was a wave of corporate fraud exposure at large

companies including WorldCom, AOL Time Warner, Global Crossing, and so on. Jeffrey

Sachs recently said that “Every Wall Street firm has paid significant fines during the past

decade for phony accounting, insider trading, securities fraud, Ponzi schemes, or outright

embezzlement by CEOs.”1 As projected in 2010 report to the nations by Association of

Certificated Fraud Examiners (ACFE), a typical organization loses 5% of its annual revenue

to fraud.

The nature of corporate fraud may differ in different survey, arguments, and media

coverage, yet, all the above reports indicate that corporate fraud is pervasively existed and

results in negative consequences. In this paper, following Dyck, Morse, and Zingales (2010),

we regard corporate fraud as misconduct behavior of firms or managers, which results in

material value loss to shareholders, or relevant stakeholders (like creditors, customers,

suppliers, and so on), and therefore leads to legal enforcement in shareholder class action

lawsuit. In particular, we pay attention to the financial implications of corporate fraud.

Previous studies explore the value implications of corporate fraud and document significant

value loss to firm and managers (Karpoff, Lee, and Martin, 2008 a&b). However, it has been

much less clear on whether and how corporate fraud affects firm’s financing costs, decisions

and polices. In this study, we aim to fill the gap by investigating the effect of corporate fraud

1 Jeffrey D. Sachs, “The Global Economy’s Corporate Crime Wave”, Project Syndicate, http://www.project-syndicate.org/commentary/the-global-economy-s-corporate-crime-wave

2

on cost of debt financing and corporate cash holdings policies because they represent the

most important external and internal financing sources.

Previous studies have proposed a straightforward connection between corporate fraud

and firms’ ability to raise additional fund. In the model of Klein and Leffler (1981) and Jarrel

and Peltzman (1985), corporate fraud damages firms’ reputation in terms of changes of terms

of trade by investors and stakeholders and therefore results in higher uncertainty about firms’

future prospects. In the framework of Kreps and Wilson (1982) and Milgrom and Roberts

(1982), corporate fraud also increases information asymmetry perceived by outsiders in the

sense that outsiders will not trust the “cheater”. The changes of terms of trade and

information asymmetry effects will make fraudulent firms more difficult and expensive to

raise additional fund. Consistent with the reputation damage argument, fraudulent firms may

be difficult to issue new debt or renegotiate existing debt agreements. The fraudulent firms

may also face costly financing cost in terms of higher interest rate and more covenant

restrictions. Taken together, the costly and restrictive external financing caused by the fraud

might force the firms to pass up valuable positive NPV projects, which would further dampen

the firm perspectives. To keep the liquidity and avoid the underinvestment issues, fraudulent

firms might rely more on internal sources of fund in terms of cash holdings accumulation to

finance future valuable investments. We therefore anticipate that fraudulent firms will

accumulate more cash and the value of cash increases associated with corporate fraud in

response to costly external finance induced by corporate fraud. Moreover, if the external

finance does become more costly after corporate fraud, fraudulent firms are expected to

become financially more constrained and as a consequence, accumulate more cash out of

cash flow (Almeida, Campello, and Weisbach, 2004).2 We therefore expect that corporate

fraud results in a higher degree of cash flow sensitivity of cash. Despite the theoretical appeal,

to our knowledge, we are among the first to directly investigate the interactions between

costly external finance and corporate cash holdings in the framework of corporate fraud.

The corporate fraud sample for this study consists of 184 fraudulent firms, which

experienced material litigation in securities class action lawsuit from 1996 until 2008 and 2 Almeida, Campello, and Weisbach (2004) model the precautionary demand of cash and find that financial constrained firms would accumulate more cash out of cash flow.

3

have issued syndicate loans both before and after corporate fraud from 1990 until 2010. In

particular, we explore the effect of corporate fraud on the cost of debt in syndicate loans

because bank lending is a particularly important source of external financing and plays an

essential role in financing firms’ investment activities (Chava and Roberts, 2008). In practice,

the syndicate loan contract commonly requires the firms to disclose litigation information for

cases that result in material changes only.3

To motivate the empirical analysis, we first explore possible channels through which

corporate fraud affects borrowing cost. As in Karpoff, Lee, and Martin (2008a), corporate

fraud mainly results in changes in two aspects. On one hand, discovery of fraud incident,

together with subsequent corporate lawsuit, conveys possible proprietary information about

fraudulent firms’ prospects and therefore leads investors to readjust estimation about

fraudulent firms’ real situation in terms of performance, corporate governance, credit risk,

and so on. On the other hand, corporate fraud damages firms’ reputation and therefore results

in high uncertainty about firms’ future prospects. Triggered by lawsuit alleged for

misconduct in some aspect, investors may question the overall big picture about fraudulent

firms’ situation and prospects. In other words, corporate fraud increases information

asymmetry perceived by outsiders. Moreover, corporate fraud damages firms’ reputation in

terms of changes of terms of trade when fraudulent firms interact with other stakeholders

such as creditors, customers, suppliers, and regulators (Karpoff, Lee, and Martin, 2008a;

Graham, Li, and Qiu, 2008). Both effects would increase the monitoring requirements and

costs of the creditors and result in a higher cost of borrowing. We therefore expect that

corporate fraud increases the cost of debt.4

We obtain supportive evidence for our hypothesis by employing a panel regression in

terms of comparing fraudulent firms’ cost of debt financing before and after fraud incident as

in Graham, Li, and Qiu (2008). Specifically, to estimate the exact magnitude effect of

corporate fraud in event study, we create the before and after fraud sample based on the

corporate fraud event trigger date instead of using the corporate fraud filing date regarding

3 As in the handbook of loan syndications and trading edited by Allison Taylor and Alicia Sansone, Page 286, “the purpose of the litigation representation is to force the borrower to identify a material litigation or threatened litigation”. 4 Since the readjustment effect is transitory, we document that the effect of corporate fraud on cost of debt is mainly due to the information asymmetry or reputation effect in long run time period.

4

the date staleness issue of fraud databases as documented in Karpoff, Koester, Lee, and

Martin (2012).5 We find that the log value of loan spread increases by 29.6 percentage after

corporate fraud. The effect is robust whenever we control a series of firm and macro

characteristics. Moreover, following Murfin (2012), we construct proxy for contract strictness

and find that probability of covenant violation increases by 15.6 percentage after corporate

fraud. The result indicates that fraudulent firms face both costly and restrictive external

financing due to high uncertainty risk induced by corporate fraud. To further confirm our

hypothesis that it is the information asymmetry problem associated with corporate fraud that

results in costly external finance, we examine the interplay of potential factor which might

capture the information asymmetry issue with corporate fraud. One particular concern is the

quality of corporate governance. Since corporate fraud results in information asymmetry

issue, the outsiders may impose more agency cost if the shareholder right is weak and

therefore can not provide sufficient monitoring for information asymmetry issue.6 In this

regard, we anticipate the costly external finance resulted by corporate fraud is more severe in

weak governance firms. Using the E-index of Bebchuk, Cohen, and Ferrell (2009), we find

that the impact of corporate fraud on cost of debt is more severe for firms with weak

governance (E-index>=4).7

To further understand how corporate fraud affects the corporate policy and financial

outcomes, we turn attentions to the effect of corporate fraud on internal sources of fund.

Based on the precautionary motive argument (Bates, Kahle, and Stulz, 2009), firms retain

cash holdings to better cope with adverse shocks when getting access to capital market is

costly. Opler, Pinkowitz, Stulz, and Williamson (1999) find that firms with greater

information asymmetry with outsiders tend to hold more cash. Faulkender and Wang (2006)

further document that the marginal value of cash increases for firms facing financing frictions

because the internal funds enable firms to invest in valuable investments projects which

5 Karpoff, Koester, Lee, and Martin (2012) document that the corporate litigation filing date lags 150 calendar days behind the initial event date for the SCAC database. 6 Previous studies also provide some argument and evidence regarding the interplay between corporate governance and cost of corporate fraud. For instance, Agrawal, Jaffe, and Karpoff (1999) argues that governance changes can recover lost reputation capital. Farber (2005) find that the fraudulent firms which take actions to improve governance can effectively reduce the cost of corporate fraud. 7 One advantage using the E-index is that variation of E-index across time is slowly (Bebchuk, Cohen, and Ferrell, 2009) and therefore can effectively avoid the endogenous issue between corporate fraud and governance changes.

5

might forgo due to costly external finance. In addition, Almeida, Campello, and Weisbach

(2004) model the precautionary demand of cash and empirically show that financial

constrained firms tend to save cash out of cash flow, while unstrained firms do not. Given the

findings that corporate fraud results in costly external financing, we anticipate that fraudulent

firms accumulate more cash and value of cash increases associated with corporate fraud

based on the precautionary motive argument. Moreover, the costly external financing

evidence also indicates that the fraudulent firms are supposed to be financial constrained after

corporate fraud. We therefore anticipate that fraudulent firms display positive cash flow

sensitivity of cash after corporate fraud.

The empirical results confirm our anticipations. In line with the findings on the cost of

debt financing, fraudulent firms accumulate more cash holdings after corporate fraud. On

average, the ratio of cash scaled by net assets increase by 2.3 percentage points. The findings

are robust for alternative measures of cash holdings. We also find that value of cash

significantly increases after corporate fraud. On average, shareholders increase their valuation

of an extra dollar of cash by $0.797 after corporate fraud. Moreover, the marginal value of

excess cash to whole firm increases by 0.564 after corporate fraud. The results suggest that

firms build cash to hedge costly external financing and thereby allow firms to invest in

positive net present value projects, which might be passed over otherwise. We further turn

attention to the effect of corporate fraud on sensitivity of cash to cash flow. Since cash flow

sensitivity of cash provides a implementable measure of the importance of financial

constrains (Almeida, Campello, and Weisbach, 2004), we identify financial constrains by

comparing the firms’ propensity to save cash out of cash flows before and after corporate

fraud. Specifically, we split the sample into two groups based on the corporate fraud filing

date and examine the cash flow sensitivity of cash, respectively. The estimations show that

the fraudulent firms display a positive cash flow sensitivity of cash for the post fraud sample

whereas do not for the before fraud sample. Our results indicate that corporate fraud is one

important firm characteristic to identify financial constrains.

To confirm the precautionary saving of cash motive in response to costly external

finance, we further pay attention to interaction effect of corporate fraud with potential factors

6

which might capture the causal effect of costly external finance on cash holdings. First, the

costly external finance punishment might be more shocked for firms in great need for

external capital. Thus, all the cash concerns in terms of cash accumulation, value of cash, and

cash flow sensitivity of cash should be more significant for fraudulent firms in industries that

are more dependent on external finance (Rajan and Zingales, 1998). Second, since cost of

corporate fraud is more severe in weak governance firms as we found in cost of debt part, the

corresponding cash concerns should be more pronounced in weak governance firms. By

constructing and ranking the industry external finance dependence proxy as in Duchin, Ozbas,

and Sensoy (2010), we find that the effect of corporate fraud on cash, value of cash and cash

flow sensitivity of cash are all more pronounced for firms in industries with high external

finance dependence. Moreover, in line with the findings in the external finance section that

cost of corporate fraud is more severe for fraudulent firms with weak governance, we find

that the effect of corporate fraud on cash accumulation, value of cash and cash flow

sensitivity of cash is more pronounced for firms with weak governance. The interplay results

further confirm the precautionary motive argument in the sense that fraudulent firms

accumulate more cash and the value of cash is more valuable when the external finance is

more costly or constrained.

Overall, our results suggest that corporate fraud increases cost of external financing in

terms of higher borrowing cost. In line with the costly external financing evidence, fraudulent

firms accumulate more cash to keep liquidity and avoid underinvestment issues. The value of

cash increases after corporate fraud. In addition, corporate fraud contributes to financial

constrains in the sense that fraudulent firms display a positive cash flow sensitivity of cash

after corporate fraud. Our result indicates that corporate fraud can have a real impact on

corporate outcomes by affecting the external financing cost and internal cash holdings.

Our study contributes to several strands of literature. First, we add to the corporate fraud

literature (Karpoff, Lee, and Martin, 2008 a&b; Karpoff and Lou, 2010; Dyck, Morse and

Zingales, 2010; Wang, Winton and Yu, 2010) by showing that corporate fraud exerts

significant impacts on firm’s financing costs, decisions and policies. The second contribution

of the paper is to show how corporate cash holdings interact with costly external finance in

7

the framework of corporate fraud. By doing this, the paper adds to corporate cash holdings

and liquidity management literature (e.g. Harford, 1999; Opler, Pinkowitz, Stulz, and

Williamson, 1999; Almeida, Campello, and Weisbach, 2004; Faulkender and Wang, 2006;

Dittmar and Mahrt-Smith, 2007; Harford, Mansi, and Maxwell, 2008) by testing the

precautionary motive argument of cash holdings in the framework of corporate fraud. In this

regard, we also contribute to the liquidity management literature (Campello, Graham, and

Harvey, 2010; Campello, Giambona, Graham, and Harvey, 2011) by showing that how

fraudulent firms manage internal sources of fund in response to costly external finance.

The remaining sections are organized as follows. Section 2 describes data construction

and summary statistics. Section 3 identities cost of corporate fraud in bank loan. Section 4

further providence evidence from fraudulent firms’ cash holdings. Section 5 concludes.

2 Data, variables, and summary statistics

2.1 Sample construction

We construct our fraudulent firms sample from Stanford Securities Class Action

Clearinghouse (SCAC). The SCAC consists of U.S. firms, which have been filed against in a

securities class action lawsuit under the provision of the federal 1033/1934 Exchange Acts

since 1996. As argued by Coffee (1986) and Choi, Nelson, and Pritchard (2009), a value

relevant fraud is less likely to emerge without a subsequent class action suit being filed.

Indeed, this dataset has been used to obtain a sample of corporate fraud recently in some

financial studies related with corporate fraud.8 One potential disadvantage of using private

class action lawsuit data is the non-meritorious (frivolous) litigation problem. Some litigation

may have nothing to do with firm’s operations. Moreover, litigation can be initiated by

lawyers’ profit-oriented incentive rather than corporate fraud (Choi, 2003). Following

previous studies (see, Dyck, Morse, and Zingales, 2010; Wang, Winton, and Yu, 2010), we

control for the frivolous lawsuit problem as follows. First, our sample is restricted to the

8 See, for example, Dyck, Morse, and Zingales (2010) on corporate fraud whistler, Wang, Winton, and Yu (2010) on fraud and business condition, Lin and Paravisini (2011) and Tian, Udell, and Yu (2012) on corporate fraud and financial intermediary monitoring.

8

period after the passage of the Private Securities Litigation Reform Act of 1995 (PSLRA),

which aims to discourage frivolous litigation (Choi, Nelson, and Pritchard, 2009; Johnson,

Nelson, and Pritchard, 2007). Second, we drop cases where the judicial review process leads

to their dismissal. Third, with respect to class action that settled, we keep only those cases

where the settlement is no less than 2 million US dollars, a threshold level of payment, which

is used to separate meritorious ones from frivolous ones as advocated by Grundfest (1995)

and Choi, Nelson, and Pritchard (2009). Fourth, we exclude the cases with the nature of

mutual fund timing and analyst malpractices, which are supposed to have nothing to do with

corporate operation.

Since our study is to examine changes about fraudulent firms’ borrowing cost and cash

holdings before and after corporate fraud, we also keep only the first lawsuit case into our

sample for firms with more than one lawsuit, following Graham, Li, and Qiu (2008).9 We

further hand check in the Dealscan database for available loan contract information from year

1990 to 2010 for these fraudulent firms. We also obtain cash holdings data and other control

variables about fraudulent firms from Compustat. We drop observations if there are missing

values for our five main firm accounting control variables in the following analysis, log firm

size, leverage, profitability, tangibility, and market to book ratio. We further exclude

fraudulent firms in financial industry.

One caveat here is how to establish the before and after fraud sample. Previous studies

create the “post fraud” indicator mainly based on the filing date in the corporate misconduct

database. For instance, Graham, Li, and Qiu (2008) construct the post event indicator based

on the date in GAO dataset. Yet, as in Karpoff, Koester, Lee, and Martin (2012), the

commonly used databases in corporate misconduct research typically misidentifies the initial

trigger date of misconduct. Without a right event date, we might bias the accurate magnitude

of the fraud event in event study. In this regard, we identify the trigger date of corporate fraud

from a series of data sources including SEC filings, Factiva, and Lexis-Nexis dataset as in

Karpoff, Koester, Lee, and Martin (2012). The date staleness issue is indeed significant.

Given our sample, the corporate litigation filing date lags 117 days behind the trigger date.

9 The estimation result has no material difference after we remove firms with more than one lawsuit.

9

The maximum lags are 1472 days. Thus, we use the corporate fraud initial trigger date

instead of the litigation filing date to construct the event indicator. Further, we remove

samples which have loan data only before or after fraud event trigger date. The final sample

consists of 1676 facility level observations and 3090 firm-year level observations by 184

fraudulent firms. Moreover, in the cash part, to pin point the accurate effect as soon as

possible, we employ estimation using the quarterly level data as in Duchin, Ozbas, and

Sensoy (2010) rather than using annual data.

2.2 Variables and summary statistics

2.2.1 Key target variables

Our loan data is from Dealscan database. Our measure for loan spread is all-in-spread

drawn variable, which is defined as the amount the borrowers pay in basis points over LIBOR

or LIBOR equivalent for the drawn portion of the loan facility. In multivariate estimation, we

use log value of loan spread as our dependent variable. Furthermore, we construct the

contract strictness proxy as in Murfin (2012) to test the probability of covenant violation

associated with corporate fraud. See appendix for details. Regarding the analysis of cash

holdings and value of cash, we mainly use cash to net assets ratio as measure of corporate

cash holdings following Opler, Pinkowitz, Stulz, and Williamson (1999), and excess return

over the fiscal year and market to book ratio as proxy for value of cash as in Faulkender and

Wang (2006) and Dittmar and Mahrt-Smith (2007), respectively. The remaining control

variables are defined in table 1.

[Insert Table 1 here]

2.2.2 Firm characteristics and univariate analysis

In panel A of table 2, we report the summary statistics of fraudulent firms’ accounting

variables at quarterly firm level. In regression of cost of debt on fraud, we control a series of

firm characteristics including log firm size, leverage, profitability, tangibility, and market to

book ratio. In analysis of corporate cash holdings, following Opler, Pinkowitz, Stulz, and

10

Williamson (1999), we control log firm size, market to book ratio, cash flow to assets ratio,

net working capital to assets ratio, capital expenditure to assets ratio, leverage, industry cash

flow risk, dividend payout dummy, R&D to sales ratio, and acquisition to assets ratio. The

detailed definition can be found in table 1.


In panel B of table 2, we construct univarate estimation by investigating changes of key

loan variables and corporate cash holdings before and after fraud in order to provide some

rough evidence for our expectations. We split the sample based on corporate fraud event

trigger date. We assign observations into after fraud sample if observations’ calendar time is

on or after corporate fraud trigger date, and before fraud sample, otherwise. As in panel B of

table 2, we find that log loan size, log loan spread and loan strictness significantly increases

after fraud. To some extent, the univariate test results support our argument: fraudulent firms

bear costly and restrictive borrowing cost due to uncertainty risk induced by corporate fraud.

Moreover, fraudulent firms accumulate more cash holdings to hedge the costly external

finance. On average, the cash to net assets ratio increases by 4.5 percentage points after

corporate fraud.

3 Cost of debt

Section 2 outlines some predications about external financing cost of corporate fraud in

bank loan contract by investigating significance of changes of loan terms around fraud event.

The univariate test is potentially misleading if fraud event is related with a series of factors

which affect cost of debt, too. In this section, we employ multivariate analysis by controlling

variables which have been well documented in previous studies as determinants of borrowing

cost. The empirical model is as follows,

Log loan spread / Contract Strictness = f(Post fraud, Loan characteristics, Borrower

characteristics, Macro variables, Industry effect, Loan purpose, Loan type, Year effect, Firm

effect)

(1)

11

The dependent variable represents log loan spread or contract strictness.10 Our key target

variable is post fraud dummy which equals one for loans issued after corporate fraud event

trigger date, and zero, otherwise. We control loan contract characteristics including log loan

size, performance pricing dummy, log maturity variable, number of lenders. We control a

series of borrower characteristics including log firm size, leverage, profitability, tangibility,

market to book ratio, and rating index. These firm factors are commonly used in previous

studies (see, for example, Graham, Li, and Qiu, 2008; Lin, Ma, Malatesta, Xuan, 2011). We

also add macro economic variables including GDP growth and credit spread variables.

3.1 Basic regression


Table 3 reports results about the impact of corporate fraud on log value of loan spread.

In column 1, we control contract characteristics, deal purpose, industry effect, and macro

economic factors. We add firm control variables including log firm size, leverage,

profitability, tangibility, market to book ratio, and rating index in column 2. In column 3, we

further control the Zscore and cash flow volatility value. We also control firm and year

effects and report heteroscedasticity-consistent robust standard errors clustered at firm level

in column 2 and 3.

Consistent with our anticipation in the sense that corporate fraud results in high cost of

debt, corporate fraud significantly and positively affect cost of borrowing. Corporate fraud

increases log loan spread by 31.2 percent at 1% significance level, ceteris paribus (column 1).

After controlling firm characteristics, the impact of the post fraud dummy on log loan spread

drops to 29.6 percent without change of significance level (column 2). This indicates that

fraudulent firm factors capture some effects of the corporate fraud but can not fully explain it.

We further add the Zscore and cash flow volatility variables into regression in column 3 and

still find significant effect of corporate fraud on loan spread. The coefficient of post fraud

dummy is 29.5 percent at 1% significance level in column 3.

10 We report the results on non-price monitoring results in appendix III for details.

12

For the remaining loan contract variables, log loan size is negatively associated with

loan spread. It may capture some scale economical lending effect or indicate that riskier firms

tend to obtain smaller size of loan with higher loan spread (Graham, Li, and Qiu, 2008). We

also find that maturity is positively related with loan spread. It is consistent with the argument

that long term loan tends to have high liquidity risk (Diamond, 1991) and has to be

compensated with high loan spread. The existence of performance pricing clause has no

statistical power of explaining variation of log loan spread. Revolver loan, i.e. credit line,

decreases loan spread significantly. We have not seen any significant impact of number of

lenders on loan spread.

With respect to firm factors, the coefficient of log firm size is negative but has no

significance (column 2 and 3). Leverage shows significant positive association with log loan

spread, indicating that high credit risk firm (high leverage) obtains a high credit spread. High

profitability may represent a low risk of default and therefore enjoy a low external financing

cost. We find a positive relationship between tangibility and log loan spread. The predication

for market to book ratio seems to be less clear. The market to book ratio can either be proxy

of growth opportunity and therefore increases credit risk or represent the additional value

over book assets left to creditor in default (Graham, Li, and Qiu, 2008) and decreases credit

risk. We find a weak positive relationship between market to book ratio and log loan spread.

This result seems to support the former argument. The S&P rating is positively related with

log loan spread. The result indicates that firm with low level of credit rating borrow at high

cost to compensate high credit risk captured by credit rating. In addition, firms with high

Zscore (low default risk) and low cash flow volatility have a low cost of debt.

For macro factors, credit spread tends to widen when economic condition is bad.

Accordingly, creditors have to be highly compensated for increased default risk in bad times.

Consistent with the argument, we find that high credit spread significantly increases loan

spread. A caveat here is that year effect seems to capture the variation of credit spread. After

we control year effect in column 2 to 3, credit spread variable is dropped due to

multicollinearity with year dummies. Finally, we find no significant result for GDP growth

and 96-02 dummy.

13

3.4 Interaction with governance on the cost of debt

We thus far find that corporate fraud results in costly external finance by investigating

the impact of corporate fraud on loan spread. To further check our hypothesis that the

increases of cost of debt are mainly in response to the information asymmetry resulted by

corporate fraud, we examine the interplay between corporate fraud and potential factor which

might capture the information asymmetry issue. We pay attention to the corporate

governance variable. If external governance can not provide sufficient monitoring for the

information asymmetry issues, outsiders might perceive more agency cost for the corporate

fraud. We therefore anticipate that the effect of corporate fraud on the cost of debt is more

severe for fraudulent firms with weak governance. One concern is that the governance quality

may change in response to corporate fraud (Agrawal, Jaffe, and Karpoff, 1999) and therefore

result in endogenous issue. In this regard, we use the E-index of Bebchuk, Cohen, and Ferrell

(2009) as proxy of corporate governance quality because the E-index is quite stable across

time (Bebchuk, Cohen, and Ferrell, 2009) and thereby avoid the endogenous problem. Indeed,

by hand-checking the changes of E-index across time in our sample, among 129 fraud cases

with E-index information, only 5 cases have a slightly improvement of governance around

the fraud incident. In particular, we recognize the firm-year observations with E-index>=4 to

be 1, indicating the weak governance sub-sample, and 0 otherwise.11 The interaction results

between corporate fraud and governance quality are reported in table 4.


In table 4, the dependent variable is the log of loan spread. Besides the post fraud

variable, we add the interaction term between post fraud and Eindex dummy variable into

regression. We control contract characteristics in column 1. We find that the cost of corporate

fraud increases by 18.0 percent at 10% significance level for the sample of weak governance

compared with the sample of good governance. The result is robust when we add the Eindex

dummy variable into regression in column 2. The results indicate that outsiders impose more

11 As in Giroud and Mueller (2011), using a cutoff of E ≥ 4 ensures that the weak governance subsample contains sufficiently many observation relative to the remaining.

14

agency cost on fraudulent firms with weak governance for their insufficient monitoring on

the information asymmetry issue.

In appendix, we also provide evidence in terms of non-price loan monitoring after

corporate fraud. See appendix III for details. To sum up, we obtain costly and restrictive

external financing cost evidence by finding that corporate fraud significantly increases loan

spread and contract strictness. This provides direct evidence of firm value loss associated

with corporate misconduct as in Karpoff, Lee, and Martin (2008a). We also find that the cost

of corporate fraud is more severe for firms with weak governance. Our results underline the

importance of governance in mitigating the cost of corporate fraud (Agrawal, Jaffe, and

Karpoff, 1999; Farber, 2005).

4 Corporate cash holdings

Section 3 outlines our basic predications on external financing cost of corporate fraud by

investigating the effect of corporate fraud on cost of debt. In this section, we aim to further

investigate the effect of corporate fraud on the corporate cash holdings. Given the findings

that corporate fraud results in costly external financing, fraudulent firms would accumulate

more cash in order to preserve liquidity and avoid underinvestment issues. We therefore

anticipate that corporate cash holdings and value of cash significantly increases associated

with corporate fraud. We further pay attention to sensitivity of cash to cash flow. Almeida,

Campello, and Weisbach (2004) model the precautionary motive of saving cash and

empirically find that financially constrained firms display a positive cash flow sensitivity of

cash, while unconstrained firms do not. Regarding the costly external financing evidence as

shown in section 3, we anticipate that fraudulent firms would become financially constrained

after corporate fraud and therefore display significantly positive cash flow sensitivity of cash.

4.1 Corporate cash holdings

Like the analysis in cost of debt, we empirically investigate the corporate cash holdings

before and after corporate fraud in panel regression. Following Opler, Pinkowitz, Stulz, and

15

Williamson (1999), the primary proxy for level of cash holdings is cash to net assets ratio.12

The key target variable is post fraud dummy. Following Opler, Pinkowitz, Stulz, and

Williamson (1999) and Bates, Kahle, and Stulz (2009), we control firm characteristics

including log firm size, market to book ratio, cash flow to assets ratio, net working capital to

assets ratio, capital expenditure to assets ratio, leverage, industry cash flow risk, dividend

payout dummy, R&D to sales ratio, and acquisition to assets ratio. All the variables are

defined in table 1. To make sure the estimation in event study is accurate, it is better when the

data frequency is we use the quarterly data in cash holdings estimation following Duchin,

Ozbas, and Sensoy (2010) rather than annual data.13 Table 5 presents the estimation results.


We find that corporate fraud significantly increases corporate cash holdings. In column

1, we control year, quarter, and industry effect. On average, the cash to net assets ratio

increases by 4.2 percent after corporate fraud at 1% significance level (column 1). In column

2, we control year, quarter and firm effects. We still find the significance level of post fraud

indicator (0.023 at 1% significance level). In unreported results, we also add a 2000s

indicator representing the 2000s time period to see whether the post fraud effect can be

captured by the increases of cash holdings in the 2000s as in Bates, Kahle, and Stulz (2009).

It turns out that the 2000s indicator variable can not explain the effect of fraud on cash.

Overall, the results confirm our anticipation that fraudulent firms accumulate more cash to

hedge costly external finance. The coefficients of remaining variables are consistent with

findings in previous studies. Similar to those documented in Opler, Pinkowitz, Stulz, and

Williamson (1999) and Bates, Kahle, and Stulz (2009), firms with large size and low growth

opportunities tend to hold less cash. Leverage ratio, net working capital to assets ratio and

acquisition to assets ratio is negatively associated with cash reserve. The industry cash flow

risk significantly increases cash holdings.

12 We obtain robust results when we use cash to total assets, log value of cash to net assets and log value of cash to sales ratio as an alternative measure following Bates, Kahle, and Stulz (2009) and Foley, Hartzell, Titman, and Twite (2007). 13 In unreported estimations, we also examine using annual data, the significance results are robust.

16

4.2 Value of cash

We further extend by investigating the value of cash associated with corporate fraud.

This helps us confirm the precautionary motive argument. Fraudulent firms accumulate more

cash to hedge costly external finance and thereby can invest in valuable project which might

pass by, otherwise. If so, we anticipate that the value of cash significantly increases

associated with corporate fraud. Indeed, Faulkender and Wang (2006) document and find that

marginal value of cash is more pronounced in financial constrained firms. We examine the

impact of corporate fraud on the value of cash from perspective of both equity holders and

the whole firm.

Following Faulkender and Wang (2006), we estimate the effect of corporate fraud on the

value of an additional dollar of cash to equity holders as follows,

, ,∆ ,

, ∗

∆ ,

,

∆ ,

,

∆ ,

,

∆ ,

,∆ ,

,

∆ ,

,

,

,,

,

,

,

,∗∆ ,

,

, ∗∆ ,

,,

(2)

where the dependent variable is the difference between firm’s stock return over fiscal year t

and the benchmark portfolio return to which stock i belongs to at the beginning of fiscal year

t. We use the 25 Fama and French portfolios formed on size and book-to-market as our

benchmark portfolios. A portfolio return is a value-weighted return based on market

capitalization within each of the 25 portfolios.14 To capture the effect of fraud on the value of

cash, we interact the post fraud dummy with the change of cash variable. For remaining

variables, see definitions in Table 1.

We further estimate the impact of corporate fraud on the value of cash to the whole

fraudulent firm. Following Pinkowitz and Williamson (2006), Dittmar and Mahrt-Smith

14 In particular, we group every firm into one of 25 size and book-to-market portfolios based on the interaction between size and book-to-market independent sorts. A firm would change the portfolio to which it belongs to during the year because the Fama and French portfolios formation date (end of each June) differs from the fiscal year end time (any month during the year). For instance, consider a firm whose fiscal end is December in year t-1, then from January to June of year t, it belongs to the portfolio based on the size and book-to-market breakpoints of year t-1, and from July to December of year t, it belongs to the portfolio based on the size and book-to-market breakpoints of year t. We then get the benchmark return by annualizing the monthly portfolio returns from the portfolio it belongs to each month.

17

(2007), and Fresard and Salva (2010), we regress market to book ratio on the level of

corporate cash holdings and a series of control variables. The estimation is as follows, ,

,

,

, ∗ ,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,,

(3)

where denotes a change in from time t-2 to t. See table 1 for definitions of variables.

The dependent variable is market value of equity at year t plus the total liabilities divided by

the net assets. We add the interaction term between the post fraud indicator and the level of

cash holdings to examine the effect of corporate fraud on the firm value of cash holdings. In

particular, following Dittmar and Mahrt-Smith (2007), we use excess cash holdings as

proxies of cash holdings.15 The calculation of excess cash holdings is described in appendix

II. We report the estimation results of value of cash in table 6.


In panel A of table 6, we report the results of value of cash to equity holders. We first

split the sample based on the fraud detection date into before fraud and post fraud sample and

report the value of cash in column 1 and 2, respectively. We find no significant effect of

changes of cash on the equity value in the before fraud sample. The result changes

dramatically for the post fraud sample. In column 2, investors value an additional dollar of

cash at 0.874$ at 1% significance level. We further run estimation using the whole sample by

adding the interaction term between post fraud and change of cash into regression in column

3.The results show that, on average, the valuation of an additional dollar of cash by equity

holders increases by 0.797 after corporate fraud. We also find that the equity value decrease

by 0.202 after corporate fraud (column 3). In column 4, we add the interactions between

change of cash and leverage and interactions between change of cash and level of cash as in

Faulkender and Wang (2006). We still find the positive association between corporate fraud

15 We get robust results by using the actual cash holdings as in Pinkowitz and Williamson (2006).

18

and value of cash. The value of cash increases by 0.656 at 1% significance level after

corporate fraud (column 4). The result is consistent with the findings in Faulkender and

Wang (2006) in the sense that marginal value of cash is more valuable for financially

constrained firms (fraudulent firms here). In addition, the level of cash holdings significantly

increases the equity value. However, the effect mainly exists for the post fraud sample. The

marginal value of cash is less valuable for shareholders in highly levered firms. The marginal

value of cash is also less valuable for firms with a high cash position. Both are consistent

with the document in Faulkender and Wang (2006). We also find that marginal increase of

earnings and marginal increase of net assets significantly increases equity value, while

marginal increase of leverage significantly decreases equity value.

We further examine the value of cash holdings from perspective of total firm valuation

in panel B of table 9. The dependent variable is the market to book value. Similar to the

estimation in panel A, we first split the sample into two groups based on the fraud detection

date and investigate the value of cash, respectively. We find no significant marginal value of

cash for the before fraud sample in column 1. By contrast, the marginal value of cash is 1.272

at 1% significance level for the post fraud sample. In the whole sample, we find that marginal

value of cash increases by 0.564 at 1% significance level after fraud (column 3). Yet,

corporate fraud itself has no significant effect on firm value. For remaining variables, high

level of earnings increase the firm value, while large increase of market value in the future

two years lower the firm value. Both are consistent with the findings in Dittmar and Mahrt-

Smith (2007).

As a whole, we find that corporate fraud significantly increases the marginal value of

cash whenever we use equity value or firm value as dependent variable. The results confirm

the argument of precautionary motive to save cash in response to costly external finance

resulted by corporate fraud. Fraudulent firms accumulate more cash to ease the external

finance constrains and thereby allow the firms to invest in positive net present value project

which might pass by, otherwise.

19

4.3 Cash flow sensitivity

In section 4.1, we obtain supportive evidence for the precautionary motive to save cash

in response to costly external finance in corporate fraud. In this section, we further examine

the sensitivity of cash holdings to cash flow associated with corporate fraud. Almeida,

Campello, and Weisbach (2004) model the liquidity demand in an imperfect capital market

and find a positive cash flow sensitivity of cash for financially constrained firms. Given the

costly external financing evidence in terms of costly price and restrictive non-price loan items

in section 3, the fraudulent firms are supposed to be more financially constrained after

corporate fraud. In this regard, we anticipate the fraudulent firms would display positive cash

flow sensitivity of cash after corporate fraud. Following Almeida, Campello, and Weisbach

(2004), we estimate the cash flow sensitivity of cash as follows,

∆ ,

, ∗ , , ,

, , ∆ , ∆ ,

,

(4)

where the dependent variable is the change of cash to assets ratio. We interact post fraud

dummy with cash flow to proxy the effect of fraud on the cash flow sensitivity of cash. See

table 1 for definitions of other control variables. We report estimations in table 7.


In table 7, we split the sample into two groups based on the fraud detection date and

estimate the cash flow sensitivity of cash before and after fraud, respectively. In column 1,

we find that no significant association between cash flow and change of cash before corporate

fraud. By contrast, we find a significantly positive cash flow sensitivity of cash after

corporate fraud in column 2. On average, the sensitivity estimate in post fraud sample is

0.266 at 1% significance level. We further estimate using the whole sample by adding the

interaction term between post fraud and cash flow variable into the regression and report

results in column 3. We find that the cash flow sensitivity increase by 0.119 after corporate

20

fraud in column 3. Since the estimation model of cash flow sensitivity of cash is essentially

an estimation of change of cash on cash flow, we also add the lag of cash holdings and lag of

change of cash holdings following the change of cash estimation in Bates, Kahle, and Stulz

(2009) and Harford, Mansi, and Maxwell (2008) into the cash flow sensitivity estimation in

column 4. The coefficient of the interaction term is 0.222 at 1% significance level, indicating

that corporate fraud incident contributes to financial constrains. One caveat here is that the

effect of post fraud on change of cash is negative after we add the interaction term between

corporate fraud and cash flow. Considering the findings in section 4.1 that corporate fraud

results in cash accumulation, it might indicate that the post fraud magnitude increase of cash

is mainly driven by the post fraud cash savings out of cash flow. In other words, the post

fraud cash accumulation might come from the post fraud saving cash out of cash flow

activities, which is indeed the model implication of Almeida, Campello, and Weisbach

(2004). Overall, the result indicates that corporate fraud incident contributes to financial

constraints. It is also consistent with the costly external finance in section 3. In addition,

consistent with Almeida, Campello, and Weisbach (2004), we also find significantly positive

Q sensitivity of cash.

4.4 Interaction with external finance dependence and governance

We obtain the precautionary saving evidence by finding that corporate fraud results in

increases of cash, value of cash, and cash flow sensitivity of cash in response to the costly

external finance. In this subsection, to further enhance the understanding of precautionary

savings of cash motive in response to the costly external finance, we investigate the

interaction effect between corporate fraud and potential factors on cash, value of cash, and

cash flow sensitivity. We first consider the industry-level external finance dependence. The

costly external finance should be more shocked for fraudulent firms in industries that are

more dependent on external finance (Rajan and Zingales, 1998). Thus, the effect of corporate

fraud on cash accumulation, the value of cash, and cash flow sensitivity of cash should be

more pronounced for firms in great need of capital. We also pay attention to the interaction

effect between corporate fraud and governance as we did in cost of debt part. Since the effect

21

of corporate fraud on cost of debt is more pronounced in firms with weak governance, we

therefore anticipate that the firms with weak governance become more financially

constrained after corporate fraud. Thus, the increases of cash and value of cash should be

more significant for firms with weak governance based on the precautionary hypothesis.

Moreover, the effect of fraud on the cash flow sensitivity of cash should be more pronounced

for the firms with weak governance (more financially constrained). In detail, we construct the

industry external finance dependence proxy following Duchin, Ozbas, and Sensoy (2010) and

create the Ex. Depend. indicator based on the industry median value of our sample (Ex.

depend.=1 for high dependence, 0, otherwise). 16 Regarding the construction of the

governance indicator, to avoid endogenous issue between governance and cash, we use the

initial value of Eindex to construct the O. Eindex dum. as in Dittmar and Mahrt-Smith (2007)

(O. Eindex dum.=1 for initial Eindex>=4, 0, otherwise).17 We estimate the interplay results

on cash, value of cash, and cash flow sensitivity of cash in table 8.


In panel A of table 8, we re-estimate the model of the determinants of cash in section 4.1

by adding post fraud interaction with finance dependence and governance indicators into

regression.18 The interplay results further confirm the precautionary hypothesis. The cash

accumulation associated with corporate fraud is more pronounced in firms with high external

finance dependence. The effect of corporate fraud on cash to net assets ratio increases by

0.030 more for the high dependence firms than the low dependence ones (column 1). This

might indicate that fraudulent firms with high external finance dependence have to

accumulate more cash to release the external financial constraints. We also find that the effect

of corporate fraud on cash accumulation is more pronounced in weak governance firms. On

average, the weak governance firms accumulate cash holdings 5.2 percent higher than the

good governance ones (column 2).

16 Following Raddatz (2006), the industry external finance dependence is as follows, first, the media value of external finance dependence is computed for each firm from 2000 to 2005. Second, the median ratio across firms in the same four-digit SIC industry is assigned to that industry. 17 As in Dittmar and Mahrt-Smith (2007) and Bebchuk, Cohen, and Ferrell (2009), the governance changes only slowly across time and the initial value is clearly exogenous to the firm value. Thus, using the initial value as in Dittmar and Mahrt-Smith (2007) can effectively reduce the endogenous issue between governance and firm value. 18 Since the industry dependence and governance indicators are constant across time for individual firm, the indicator impact is absorbed by the firm effects.

22

We further estimate the interplay effect on the value of cash to equity holders in panel B

of table 8. As shown in panel B of table 8, the effect of corporate fraud on value of cash is

more pronounced in high external finance dependence firms (coefficient of interaction term

among post fraud, change of cash, and Ex. Depend. in column 1). In column 2, we also find

that the increase of value of cash associated with corporate fraud in weak governed firms is

higher than that in good governed ones. The magnitude increase of value of cash associated

with corporate fraud in weak governance firms is 2.162 higher than the effect in good

governance ones (coefficient of interaction term among post fraud, change of cash, and O.

Eindex dum. in column 2). This might occur because the weak governance firms become

more financially constrained after corporate fraud and therefore the value of cash is more

pronounced (Faulkender and Wang, 2005). In addition, we also find that the value of cash

decrease by 2.326 at 1% significance level for the weak governance firms (coefficient

between change of cash and O. Eindex dum. in column 2). This is consistent with Dittmar

and Mahrt-Smith (2007) in the sense that the value of cash is higher in good governance

firms.

Finally, we examine whether the interaction effect holds for the cash flow sensitivity of

cash. Panel C of table 8 reports the results. In column 1, the cash flow sensitivity of cash

associated with corporate fraud in high dependence firms increases 0.028 more than the firms

in low dependence ones (the coefficient of interaction term among post fraud, cash flow, and

Ex. depend.).19 In column 2, we also find that the cash flow sensitivity of cash associated

with corporate fraud is more pronounced for the weak governance firms. The coefficient of

the interaction term between corporate fraud and cash flow is 0.182 at 5% significance level.

The effect further increases by 0.997 at 1% significance level for the weak governance firms.

We therefore conclude that weak governance firms become more financially constrained after

corporate fraud based on the findings in cash flow sensitivity estimation. This is consistent

19 The coefficient in this estimation is not significant. In unreported results, we split the sample into two groups based on the industry median external finance dependence degree, and estimate the effect of corporate fraud on the cash flow sensitivity in the high and low dependence groups, respectively. We find that, first, the effect of corporate fraud on cash flow sensitivity are both significant for the two groups, second, the magnitude difference between the two groups are statistically not equal to zero. Consistent with our anticipation, the magnitude effect for the high dependence group is higher than the low dependence group.

23

with the findings in section 3.2 in the sense that weak governance firms face a higher external

cost of debt after corporate fraud.

Overall, our results suggest that fraudulent firms retain more cash in response to costly

external finance. It empirically supports the precautionary motive argument (Opler,

Pinkowitz, Stulz, and Williamson, 1999). Consistent with the argument, the value of cash

significantly increases after fraud. We also show a significantly positive cash flow sensitivity

of cash after corporate fraud. The examinations show that fraudulent firms become more

financially constrained due to costly and restrictive external financing. We further confirm

the precautionary hypothesis by examining the interaction effect of fraud with external

finance dependence and governance on the cash, value of cash, and cash flow sensitivity of

cash. Consistent with the findings in section 3.4 that external cost of debt associated with

corporate fraud is more severe for firms with weak governance, we also find that the effect of

corporate fraud on cash, value of cash, and cash flow sensitivity of cash is more pronounced

for firms with weak governance.

5 Conclusion

Corporate fraud is one important issue in corporate finance studies. The fraud incident

can result in reputation damage for both firm and managers (Karpoff, Lee, and Martin, 2008

a&b). In this paper, we examine channels through which corporate fraud affects corporate

values. Through comparing the differences of cost of debt before and after corporate fraud,

we find that corporate fraud significantly increases the external financing cost. We also find

that fraudulent firms face more covenant restrictions. In line with the costly and restrictive

external finance, fraudulent firms retain more cash to better cope with adverse shocks.

Consistent with the precautionary motive argument, shareholders and outsiders value more

for an additional dollar of cash after corporate fraud. Additionally, we document that

corporate fraud contribute to financial constrains by finding that fraudulent firms save more

cash out of cash flow.

24

Reference

Agrawal, A., Jaffe, J., Karpoff J., 1999. Management Turnover and Governance Changes

Following the Revelation of Fraud. Journal of Law and Economics, April 1999, V. 42(1), 23-56.

Almeida, H., Campello, M., Weisbach, M., 2004. The Cash Flow Sensitivity of Cash. The

Journal of Finance, 59: 1777–1804. Bates, W., Kahle, M., Stulz, R., 2009. Why Do U.S. Firms Hold So Much More Cash than

They Used To?. Journal of Finance 64, 1985–2021. Bebchuk, L., Cohen, A., Ferrell, A., 2009. What Matters in Corporate Governance?. Review

of Financial Studies, 22(2): 783-827. Campello, M., Graham, J., Harvey, C., 2010. The Real Effects of Financial Constraints:

Evidence from a Financial Crisis. Journal of Financial Economics 97, 470-487. Campello, M., Giambona, E., Graham, J., Harvey, C., 2011. Liquidity Management and

Corporate Investment during a Financial Crisis. Review of Financial Studies 24, 1944-1979.

Chava, S., Roberts, M., 2008. How Does Financing Impact Investment? The Role of Debt

Covenants. Journal of Finance 63, 2085-2121. Choi, S., 2003. The Evidence on Securities Class Actions. Vanderbilt Law Review 56, 1466-

1525. Choi, S., Nelson, K., Pritchard, A., 2009. The Screening Effect of the Private Securities

Litigation Reform Act. Journal of Empirical Legal Studies 6, 35–68. Coffee, J., 1986. Understanding the plaintiff's attorney: The implications of economic theory

for private enforcement of law through class and derivative actions. Columbia Law Review 86, 669–727.

Diamond, D., 1991. Debt Maturity Structure and Liquidity Risk. Quarterly Journal of

Economics, MIT Press, vol. 106(3), pages 709-37, August. Dittmar, A., Mahrt-Smith, J., 2007. Corporate governance and the value of cash holdings.

Journal of Financial Economics 83, 599-634. Duchin, R., Ozbas, O., Sensoy, B., 2010. Costly External Finance, Corporate Investment, and

the Subprime Mortgage Credit Crisis. Journal of Financial Economics, Volume 97, Issue 3, Pages 418-435.

Dyck, A., Morse, A., Zingales, L., 2010. Who Blows the Whistle on Corporate Fraud?.

Journal of Finance, 65, 2213–2253. Farber, D., 2005. Restoring Trust after Fraud: Does Corporate Governance Matter?. The

Accounting Review: April 2005, Vol. 80, No. 2, pp. 539-561.

25

Faulkender, M., Wang, R., 2006. Corporate Financial Policy and the Value of Cash. Journal

of Finance, 61, 1957–1990. Foley, C., Hartzell, C., Titman, S., Twite, G., 2007. Why do Firms Hold so much Cash? A

Tax-based Explanation. Journal of Financial Economics, Elsevier, vol. 86(3), pages 579-607, December.

Fresard, L., Salva, C., 2010. The value of excess cash and corporate governance: Evidence

from US cross-listings. Journal of Financial Economics 98, 359-384. Giroud, X., Mueller, H., 2011. Corporate Governance, Product Market Competition, and

Equity Prices. The Journal of Finance, 66: 563–600. Graham, J., Li, S., Qiu, J., 2008. Corporate misreporting and bank loan contracting. Journal

of Financial Economics 89, 44-61. Grundfest, J., 1995. Why disimply? Harvard Law Review 108, 740–741. Harford, J., 1999. Corporate Cash Reserves and Acquisitions. Journal of Finance 54, 1969-

1997. Harford, J., Mansi, S., Maxwell, W., 2008. Corporate governance and firm cash holdings in

the US. Journal of Financial Economics 87, 535-555. Jarrell, G., Peltzman, S., 1985. The Impact of Product Recalls on the Wealth of Sellers.

Journal of Political Economy 93, 512–536. Johnson, M., Nelson, K., Pritchard, A., 2007. Do the Merits Matter More? Class actions

under the Private Securities Litigation Reform Act. Journal of Law, Economics, and Organization 23, 627-652.

Karpoff, J., Koester, A., Lee, S., Martin, G., 2012. A Critical Analysis of Databases Used in

Financial Misconduct Research. Mays Business School Research Paper No. 2012-73; Georgetown McDonough School of Business Research Paper No. 2012-11.

Karpoff, J., Lee, D., Martin, G., 2008a. The Cost to Firms of Cooking the Books. Journal of

Financial and Quantitative Analysis 43, 581-611. Karpoff, J., Lee, D., Martin, G., 2008b. The Consequences to Managers for Financial

Misrepresentation. Journal of Financial Economics 88, 193-215. Karpoff, J., Lou, X., 2010. Short Sellers and Financial Misconduct, Journal of Finance 65,

1879-1913. Klein, B., Leffler, K., 1981. The Role of Market Forces in Assuring Contractual Performance.

Journal of Political Economy 89, 615–641. Kreps, D., Wilson, R., 1982. Reputation and imperfect information. Journal of Economic

Theory, Elsevier, vol. 27(2), pages 253-279, August.

26

Lin, C., Ma, Y., Xuan, Y., 2011. Ownership Structure and Financial Constraints: Evidence from a Structural Estimation. Journal of Financial Economics 102, 416-431.

Lin, C., Ma, Y., Malatesta, P., Xuan, Y., 2011. Ownership Structure and the Cost of

Corporate Borrowing. Journal of Financial Economics 100, 1-23. Lin, H., Paravisini, D., 2011. Delegated Monitoring of Fraud: The Role of Non-Contractual

Incentives. Working paper, Columbia University GSB. Milgrom, P., Roberts, J., 1982. Predation, Reputation, and Entry Deterrence. Journal of

Economic Theory 27, 280-312. Murfin, J., 2012. The Supply-Side Determinants of Loan Contract Strictness. The Journal of

Finance, 67: 1565–1601. Opler, T., Pinkowitz, L., Stulz, R., Williamson, R., 1999. The Determinants and Implications

of Corporate Cash Holdings. Journal of Financial Economics 52, 3-46. Pinkowitz, L., Williamson, R., 2006. What is a Dollar Worth? The Market Value of Cash

Holdings. Working Paper, Georgetown University, Washington, DC, unpublished. Raddatz, C., 2006. Liquidity Needs and Vulnerability to Financial Underdevelopment.

Journal of Financial Economics, Elsevier, vol. 80(3), pages 677-722. Rajan, R., Zingales L., 1998. Financial Dependence and Growth. The American Economic

Review , Vol. 88, No. 3, pp. 559-586. Tian, X., Udell, G., Yu, X., 2012. Disciplining Delegated Monitors: The Consequences of

Failing to Prevent Fraud. Sixth Singapore International Conference on Finance 2012 Paper.

Wang, T., Winton, A., Yu, X., 2010. Corporate Fraud and Business Conditions: Evidence

from IPOs. The Journal of Finance 65, 2255–2292.

27

Table 1. Definitions of variables This table provides detailed definitions for all the variables used in the paper.

Variables names Variable definitions Sources Loan characteristics Log loan size Log value of loan amount (facility level), measured in millions of US dollars Dealscan Log maturity Log value of maturity (facility level), measured in months Dealscan Log loan spread Log value of loan spread. Loan spread is the all-in-spread drawn, defined as the

amount the borrowers pays in basis points over LIBOR or LIBOR equivalent for the drawn portion of the loan facility

Dealscan

Price grid A dummy variable equal to one if the loan facility has at least a performance pricing clause, zero, otherwise

Dealscan

Revolver A dummy variable equal to one for the credit line type loan, zero, otherwise DealscanNum. of lenders The number of lenders participating into loan facility syndication. DealscanFirm characteristics Log firm size Log value of firm assets measured in millions of US dollars Compustat Leverage (Long-term debt + debt in current liabilities)/firm assets Compustat Profitability Net income/ firm assets Compustat Tangibility Net property, plant, and equipment / firm assets Compustat Market to book (Market value of equity + Book value of debt)/ firm assets. Market value of equity

equals price per share times total number of shares outstanding. Book value of debt equals total assets minus book value of equity

Compustat

Cash flow volatility Standard deviation of yearly cash flows from operations over the five fiscal years prior to the loan initiation scaled by firm assets

Compustat

S&P ratings S&P firm credit ratings are converted to an index from 1 to 7 as follows: 1=Aaa, 2=Aa, 3=A, 4=Bbb, 5=Bb, 6=B or worse, and 7=no rating

Compustat

Zscore Modified Altman’s (1968) Z-score= (1.2*working capital +1.4*retained earnings +3.3*EBIT +0.999*sales)/firm assets. Following Graham, Li, and Qiu (2008), we exclude the ratio of market to book ratio in Zscore calculation because Market to book ratio itself also enters regression estimations.

Compustat

Cash holdings Holdings of cash and marketable securities/firm assets Compustat Lag of cash holdings Lagged value of cash holdings Compustat Lag of change of cash holdings

Change of cash holdings from year t-2 to t-1 Compustat

Cash flow Earnings after interest, dividends, and taxes but before depreciation scaled by firm assets

Compustat

NWC Following Bates, Kahle, and Stulz (2009), we subtract cash from net working capital and use the remaining scaled by firm assets

Compustat

Change of NWC Change of NWC from year t-1 to t Compustat Capex Capital expenditure to firm assets. For the quarterly data calculation, because capital

expenditure is reported on a year-to-date basis in quarterly financial statements, we subtract the previous quarter’s capital expenditure from the current quarter’s capital expenditure for fiscal quarters 2, 3, and 4 following Duchin, Ozbas, and Sensoy (2010).

Compustat

Dividend dum. Equal to one in years or quarters a firm pays a common dividend, zero otherwise CompustatR&D/sales R&D/sales, and set equal to zero if R&D is missing Compustat Acquisition activity Acquisitions/firm assets Compustat Industry sigma Standard deviation of industry cash flow to firm assets. The calculation method

follows Bates, Kahle, and Stulz (2009). For each firm-year-quarter observations, we compute the standard deviation of cash flow to assets for the previous 16 quarters. We require at least three observations. We then average the firm cash flow standard deviation each quarter across each two-digit SIC code

Compustat

Change of cash The change of cash holdings from fiscal year t-1 to t divided by market value of equity in year t-1

Compustat

Change of earnings The change of earnings from fiscal year t-1 to t divided by market value of equity in year t-1. Earnings are calculated earnings before extraordinary items plus interest, deferred tax credits, and investment tax credits

Compustat

(Continued on the next page)

28

Table 1. Definitions of variables (Continued) Change of net assets The change of net assets from fiscal year t-1 to t divided by market value of equity in

year t-1. Net assets are calculated as firm total assets minus cash and marketable securities

Compustat

Change of R&D The change of R&D expenditures from fiscal year t-1 to t divided by market value of equity in year t-1. R&D expenditures are set to zero if missing

Compustat

Change of interest expense

The change of interest expense from fiscal year t-1 to t divided by market value of equity in year t-1

Compustat

Change of div The change of common dividends paid from fiscal year t-1 to t divided by market value of equity in year t-1

Compustat

Cash Holdings of cash and marketable securities in fiscal year t-1divided by market value of equity in year t-1

Compustat

Market leverage Market debt ratio, calculated as total debt over the sum of total debt and the market value of equity

Compustat

Net financing Net financing during fiscal year t divided by the market value of equity in year t-1. Net financing is total equity issuance minus repurchase plus debt issuance minus debt redemption

Compustat

Excess cash/NA The excess cash divided by net assets. The excess cash is calculated as cash holdings minus the optimal cash level estimated from the appendix II

Compustat

Earnings/NA Earnings divided by net assets. Earnings are calculated earnings before extraordinary items plus interest, deferred tax credits, and investment tax credits

Compustat

2( L2) Earnings/NA The two-year future (lagged) change of earnings divided by net assets Compustat R&D/NA R&D expenditures divided by net assets Compustat

2( L2) R&D/NA The two-year future (lagged) change of R&D expenditures divided by net assets Compustat Dividends/NA Common dividends paid divided by net assets Compustat

2( L2) Dividends/NA The two-year future (lagged) change of Common dividends paid divided by net assets Compustat Interests/NA Interests expenses divided by net assets Compustat 2 L2 Interests/NA The two-year future (lagged) change of interests expenses divided by net assets Compustat 2 L2 NA/NA The two-year future (lagged) change of net assets divided by net assets Compustat 2Marketvalue/NA The two-year future change of market value divided by net assets. Market value is

calculated as price times shares plus total liabilities Compustat

Shortdebt Debt in current liabilities scaled by firm assets Compustat Changeofshortdebt Change of short debt from t-1 to t Compustat Macrosituations GDPgrowth Annual GDP growth rate of US The Federal

Reserve Board of Governors

Creditspread The difference between Moody’s Seasoned Baa Corporate Bond Yield and Aaa Corporate Bond Yield

The Federal Reserve Board of Governors

96‐02dummy A dummy variable that equals one if loan issue between year 1996 and 2002, zero otherwise

Author calculation

Others Postfraud A dummy variable that equals one if loan or firm observation is on or after fraud

trigger event date, zero otherwise SCAC

Eindexdum. A dummy variable that equals one if firm-year Eindex no less than 4, zero otherwise Bebchuk, Cohen, and Ferrell (2009)

O.Eindexdum. A dummy variable that equals one if firm original E-index no less than 4, zero otherwise

Bebchuk, Cohen, and Ferrell (2009)

Externalfinancedependence

(Capital expenditures-funds from operations)/capital expenditures. Funds from operations is replaced as the sum of the followings variables: income before extraordinary items, depreciation and amortization, deferred taxes, equity in net loss/earnings, sale of property, plant, and equipment and investments-gain/loss, and funds from operations – other, if it is missing

Compustat

Ex.Depend. A dummy variable that equals one if firm’s industry external finance dependence is greater than the industry median, zero, otherwise.

Author calculation

29

Table 2. Summary Statistics on fraudulent firms and their loans In this table, panel A presents quarterly firm level statistics on fraudulent firm ratios including number of observations (N), mean and standard deviation (sd). Panel B presents mean value for loans originated before and after fraud, respectively. We also report the difference value (mean) and t statistics between variables after fraud and before fraud. Definitions of all variables are reported in Table 1. Significance at 10%, 5%, and 1% level is indicated by *, **, and ***, respectively.

Panel A Firm variables

N mean sd Log firm Size 11746 7.111 1.974 Market to book 11176 2.071 1.205 Leverage 11270 0.311 0.193 Cash flow 10865 0.037 0.029 NWC 11335 0.068 0.161 Capex 10772 0.017 0.015 Dividend dum. 12421 0.166 0.372 R&D/sales 12176 0.023 0.044 Acquisitions 10335 0.005 0.012 Industry sigma 12421 0.310 0.384

Panel B Univariate analysis

After fraud Before fraud Difference

mean mean mean T Statistics

Log loan size 19.157 18.888 0.270*** 3.39

Log loan spread 4.990 4.678 0.312*** 6.18

Strictness 0.618 0.489 0.128*** 4.28

Cash holdings 0.149 0.104 0.045*** 15.27

30

Table 3. The effect of corporate fraud on loan pricing This table presents the regression results on the effect of corporate fraud on loan pricing. The dependent variable is the log loan spread for a single bank loan (facility level). The post fraud dummy is defined to be equal 1 for facilities which originate after corporate fraud and 0, otherwise. Definitions of all the other variables are reported in Table 1. Heteroscedasticity-consistent robust standard errors are reported in brackets in column 1. Heteroscedasticity-consistent robust standard errors clustered at firm level are reported in brackets in column 2 and 3. Significance at the 10%, 5%, and 1% level is indicated by *, **, and ***, respectively. (1) (2) (3)

Post fraud 0.312 0.296 0.295

[0.041]*** [0.060]*** [0.070]***

Log loan size -0.282 -0.071 -0.061

[0.016]*** [0.021]*** [0.020]***

Log maturity 0.163 0.065 0.055

[0.033]*** [0.028]** [0.031]*

Price grid -0.024 -0.022 -0.034

[0.040] [0.037] [0.033]

Revolver -0.234 -0.121 -0.117

[0.040]*** [0.027]*** [0.026]***

Num. of lenders -0.002 -0.002 -0.001

[0.002] [0.002] [0.001]

Log firm size -0.047 -0.035

[0.063] [0.071]

Leverage 0.635 0.911

[0.197]*** [0.246]***

Profitability -1.329 -1.239

[0.319]*** [0.531]**

Tangibility 0.545 0.875

[0.314]* [0.327]***

Market to book 0.030 0.053

[0.017]* [0.028]*

S&P ratings 0.191 0.177

[0.030]*** [0.030]***

Zscore -0.122

[0.057]**

Cash flow volatility 1.750

[0.890]*

GDP growth 0.038 0.071 -0.077

[0.023] [0.129] [0.057]

Credit spread 0.831

[0.125]***

96-02 dummy -0.032

[0.044]

Constant 8.883 4.719 2.866

[0.372]*** [0.778]*** [0.772]***

Deal purpose Yes Yes Yes

Industry Yes Yes Yes

Year No Yes Yes

Firm No Yes Yes

Observations 1366 1366 1134

Adjusted R-squared 0.496 0.825 0.854

31

Table 4. The interaction effect of corporate fraud and governance on loan pricing This table presents the regression results on the interaction effect of corporate fraud and governance on loan pricing. The dependent variable is the log loan spread for a single bank loan (facility level). The post fraud dummy is defined to be equal 1 for facilities which originate after corporate fraud and 0, otherwise. Definitions of all the other variables are reported in Table 1. Heteroscedasticity-consistent robust standard errors clustered at firm level are reported in brackets. Significance at the 10%, 5%, and 1% level is indicated by *, **, and ***, respectively. (1) (2)

Post fraud 0.198 0.191

[0.084]** [0.086]**

Post fraud * Eindex dum. 0.180 0.240

[0.099]* [0.122]*

Eindex dum. -0.113

[0.113]

Log loan size -0.075 -0.074

[0.025]*** [0.025]***

Log maturity 0.081 0.081

[0.038]** [0.038]**

Price grid -0.021 -0.023

[0.046] [0.046]

Revolver -0.096 -0.095

[0.036]*** [0.036]***

Num. of lenders -0.003 -0.003

[0.002] [0.002]

Log firm size 0.051 0.051

[0.086] [0.085]

Leverage 0.892 0.906

[0.312]*** [0.313]***

Profitability -2.099 -2.113

[0.652]*** [0.657]***

Tangibility 0.744 0.743

[0.350]** [0.350]**


[0.045] [0.045]

S&P ratings 0.236 0.234

[0.029]*** [0.030]***

GDP growth 0.025 0.026

[0.069] [0.069]

Constant 3.801 3.801

[0.783]*** [0.776]***

Deal purpose Yes Yes

Industry Yes Yes

Year Yes Yes

Firm Yes Yes

Observations 812 812

Adjusted R-squared 0.853 0.853

32

Table 5. The effect of corporate fraud on corporate cash holdings The table presents the regression results on the effect of corporate fraud on corporate cash holdings. The dependent variable is the cash to net assets ratio. The post fraud dummy is defined to be equal 1 for firm observations which are on or after corporate frauds and 0, otherwise. Definitions of all the other variables are reported in Table 1. The estimation is reported year, quarter, and industry effects in column 1, and is reported year, quarter, and firm fixed effects in column 2. Heteroscedasticity-consistent robust standard errors is reported in brackets in column 1 and heteroscedasticity-consistent robust standard errors clustered at firm level is reported in brackets in column 2. Significance at the 10%, 5%, and 1% level is indicated by *, **, and ***, respectively. (1) (2)


[0.004]*** [0.009]***


[0.001]*** [0.006]***


[0.002]*** [0.005]***

Leverage -0.095 -0.008

[0.010]*** [0.026]

Cash flow -0.393 -0.049

[0.062]*** [0.132]

NWC -0.133 -0.063

[0.013]*** [0.040]

Capex 0.292 0.099

[0.111]*** [0.209]

Dividend dum. 0.013 0.022

[0.003]*** [0.011]*

R&D/sales 0.559 0.479

[0.047]*** [0.171]***

Acquisition activity -0.405 -0.172

[0.105]*** [0.113]

Industry sigma 0.020 0.009

[0.005]*** [0.010]


[0.013]*** [0.045]***



33

Table 6. The effect of corporate fraud on value of cash The table presents the regression results on the effect of corporate fraud on the value of cash holdings. In panel A, the dependent variable is the abnormal return using the size and book to market 25 portfolios as benchmark to proxy the value to equity holders. In panel B, the dependent variable is the market to book ratio as a measure of firm value. The post fraud dummy is defined to be equal 1 for firm observations which are on or after corporate frauds and 0, otherwise. Definitions of all the other variables are reported in Table 1. All the estimations are reported firm and year fixed effects. Heteroscedasticity-consistent robust standard errors clustered at firm level are reported in brackets. Significance at the 10%, 5%, and 1% level is indicated by *, **, and ***, respectively.

Panel A: The effect of corporate fraud on value of cash to equity holders (1) (2) (3) (4)

Before After Whole Whole

Post fraud -0.202 -0.225

[0.056]*** [0.053]***

Post fraud * change of cash 0.797 0.656

[0.280]*** [0.192]***

Change of cash 0.414 0.874 -0.166 0.719

[0.273] [0.207]*** [0.298] [0.231]***

Change of earnings 0.342 0.041 0.100 0.096

[0.116]*** [0.041] [0.033]*** [0.027]***

Change of net assets 0.082 0.124 0.073 0.084

[0.050] [0.044]*** [0.042]* [0.044]*

Change of R&D -1.547 -1.347 -0.809 -0.736

[2.132] [0.905] [0.872] [0.870]

Change of interest expense 3.557 -1.194 -0.706 -0.311

[1.776]** [0.447]*** [0.487] [0.445]

Change of div 1.767 -1.039 0.552 0.805

[3.216] [1.593] [1.740] [1.600]

Cash 0.070 0.403 0.161 0.253

[0.181] [0.129]*** [0.087]* [0.071]***

Market leverage -0.594 -0.114 -0.268 -0.245

[0.278]** [0.101] [0.098]*** [0.101]**

Net financing -0.009 -0.205 -0.054 -0.067

[0.161] [0.115]* [0.080] [0.082]

Change of cash * cash -0.319

[0.074]***

Change of cash * market leverage -0.431

[0.150]***

Constant 0.385 -0.196 0.284 0.222

[0.204]* [0.090]** [0.190] [0.173]

Observations 455 725 1180 1180

Adjusted R-squared 0.131 0.263 0.133 0.180

34

Table 6. The effect of corporate fraud on value of cash (Continued)

Panel B: The effect of corporate fraud on value of cash to total firm

(1) (2) (3)

Before After Whole

Post fraud 0.424

[0.273]

Post fraud * Excess cash/NA 0.564

[0.275]**

Excess cash/NA 0.473 1.272 0.437

[0.425] [0.413]*** [0.288]

Earnings/NA 6.810 3.903 5.341

[1.668]*** [0.946]*** [0.996]***

L2 Earnings/NA 2.404 -0.037 0.388

[0.795]*** [0.528] [0.572]

2 Earnings/NA 0.679 1.456 1.456

[0.457] [0.428]*** [0.371]***

R&D/NA -14.426 9.297 -1.543

[10.603] [5.643] [5.313]

L2 R&D/NA 1.572 -17.332 -13.934

[16.529] [9.468]* [9.739]

2 R&D/NA -3.306 12.566 6.167

[7.812] [6.430]* [6.672]

Dividends/NA -0.375 -7.897 5.602

[14.930] [10.403] [12.784]

L2 Dividends/NA -38.874 -36.315 -19.700

[23.550] [16.435]** [13.353]

2 Dividends/NA -43.354 -2.205 -3.105

[17.090]** [12.483] [12.815]

Interests/NA 6.166 7.472 3.964

[6.270] [6.103] [4.483]

L2 Interests/NA 0.830 0.609 0.292

[5.808] [3.735] [3.250]

2 Interests/NA 5.420 1.218 1.079

[4.668] [5.042] [3.341]

L2 NA/NA -0.313 -0.046 0.023

[0.253] [0.185] [0.135]

2 NA/NA 0.704 0.479 0.480

[0.126]*** [0.121]*** [0.081]***

2 Market value/NA -0.274 -0.327 -0.310

[0.060]*** [0.046]*** [0.037]***

Constant 1.894 1.932 1.495

[0.660]*** [0.392]*** [0.519]***

Observations 248 343 591


35

Table 7. The effect of corporate fraud on cash flow sensitivity of cash This table presents impact of corporate fraud on cash flow sensitivity of cash. The dependent variable is the changes of cash to assets ratio. The post fraud dummy is defined to be equal 1 for firm observations which are on or after corporate frauds and 0, otherwise. Definitions of all the other variables are reported in Table 1. All estimations include year, quarter, and firm fixed effects and report heteroskesticity-consistent robust standard errors clustered at firm level in brackets. Significance at the 10%, 5%, and 1% level is indicated by *, **, and ***, respectively. (1) (2) (3) (4)

Before After Whole Whole

Post fraud -0.007 -0.009

[0.002]*** [0.002]***

Lag of cash holdings -0.203

[0.015]***

Lag of change of cash holdings -0.080

[0.015]***

Post fraud * cash flow 0.119 0.222

[0.052]** [0.058]***

Cash flow 0.047 0.266 0.090 -0.053

[0.069] [0.057]*** [0.045]** [0.052]

Market to book 0.006 0.000 0.003 0.007

[0.002]*** [0.002] [0.001]*** [0.001]***

Log firm size -0.001 -0.001 -0.002 -0.004

[0.002] [0.002] [0.001]* [0.001]***

Capex -0.637 -0.629 -0.550 -0.566

[0.107]*** [0.107]*** [0.067]*** [0.072]***

Acquisition activity -0.692 -1.254 -0.902 -0.747

[0.096]*** [0.134]*** [0.080]*** [0.074]***

Change of NWC -0.562 -0.578 -0.571 -0.498

[0.049]*** [0.048]*** [0.037]*** [0.033]***

Change of short debt -0.723 -0.466 -0.593 -0.506

[0.083]*** [0.058]*** [0.051]*** [0.046]***

Constant 0.009 0.020 0.013 0.052

[0.012] [0.013] [0.006]** [0.008]***

Observations 3784 4961 8745 8648

Adjusted R-squared 0.175 0.254 0.201 0.305

36

Table 8. The interaction effect of corporate fraud with governance and external finance

dependence on cash, value of cash, and cash flow sensitivity The table presents the regression results on the interaction effect of corporate fraud with governance and external finance dependence on cash, value of cash, and cash flow sensitivity in panel A, panel B, and panel C, respectively. In panel A, the dependent variable is cash to net assets ratio. In panel B, the dependent variable is abnormal return using the size and book to market 25 portfolios as benchmark to proxy the value to equity holders. In panel C, the dependent variable is the change of cash to assets ratio. The post fraud dummy is defined to be equal 1 for firm observations which are on or after corporate frauds and 0, otherwise. Definitions of all the other variables are reported in Table 1. All the estimations are reported firm and year fixed effects. Quarter effect is also controlled in estimations of Panel A and Panel C. Heteroscedasticity-consistent robust standard errors clustered at firm level are reported in brackets. Significance at the 10%, 5%, and 1% level is indicated by *, **, and ***, respectively.

Panel A: The interaction effect on cash (1) (2)


[0.012] [0.011]

Post fraud * Ex. depend. 0.030

[0.014]**

Post fraud * O. Eindex dum. 0.052

[0.022]**


[0.006]*** [0.010]***


[0.005]*** [0.006]***

Leverage -0.006 0.029

[0.027] [0.038]

Cash flow -0.053 0.128

[0.132] [0.202]

NWC -0.063 -0.120

[0.039] [0.052]**

Capex 0.110 -0.041

[0.208] [0.316]

Dividend dum. 0.021 0.026

[0.011]* [0.016]

R&D/sales 0.479 0.253

[0.169]*** [0.115]**


[0.115]* [0.137]**

Industry sigma 0.005 0.026

[0.010] [0.015]*


[0.045]*** [0.076]***



37


dependence on cash, value of cash, and cash flow sensitivity (Continued)

Panel B: The interaction effect on value of cash (1) (2)

Post fraud -0.260 -0.219

[0.060]*** [0.068]***


[0.072]

Ex. depend. * change of cash -0.358

[0.337]

Post fraud * O. Eindex dum. -0.193

[0.109]*

Change of cash * O. Eindex dum. -2.326

[0.666]***

Post fraud * change of cash 0.275 0.528

[0.241] [0.353]

Post fraud * change of cash * Ex. depend. 0.862

[0.355]**

Post fraud * change of cash * O. Eindex dum. 2.162

[0.806]***

Change of cash 0.885 1.574

[0.227]*** [0.480]***

Change of earnings 0.110 0.096

[0.029]*** [0.091]

Change of net assets 0.084 0.243

[0.047]* [0.061]***

Change of R&D -0.605 0.517

[0.892] [1.591]

Change of interest expense -0.290 -3.030

[0.411] [0.882]***

Change of div 0.417 1.677

[1.609] [3.216]

Cash 0.278 0.947

[0.072]*** [0.190]***

Market leverage -0.248 -0.113

[0.101]** [0.150]

Net financing -0.075 -0.317

[0.093] [0.196]

Change of cash * cash -0.299 0.195

[0.075]*** [0.154]

Change of cash * market leverage -0.509 -1.221

[0.175]*** [0.451]***


[0.176] [0.087]



38


dependence on cash, value of cash, and cash flow sensitivity (Continued)

Panel C: The interaction effect on cash flow sensitivity (1) (2)

Post fraud -0.011 -0.005

[0.003]*** [0.003]

Lag of cash holdings -0.204 -0.202

[0.015]*** [0.020]***

Lag of change of cash holdings -0.079 -0.051

[0.015]*** [0.021]**


[0.003]

Post fraud * O. Eindex dum. -0.030

[0.009]***

Post fraud * cash flow * Ex. depend. 0.028

[0.087]

Post fraud * cash flow * O. Eindex dum. 0.997

[0.255]***

Post fraud * cash flow 0.208 0.182

[0.075]*** [0.085]**

Cash flow -0.057 0.044

[0.051] [0.079]


[0.001]*** [0.001]**


[0.001]*** [0.002]**

Capex -0.566 -0.503

[0.071]*** [0.086]***


[0.076]*** [0.106]***

Change of NWC -0.499 -0.561

[0.033]*** [0.045]***

Change of short debt -0.506 -0.441

[0.046]*** [0.065]***


[0.008]*** [0.013]***



39

Appendix I Measuring Contract Strictness

As in Murfin (2012), we construct the contract strictness proxy to capture the number,

slackness, scale, and covariance of covenants. Consider contracts with 1 vector of

financial ratios which receives an dimensional shock, migrating to ,

~ ,

(A.1)

Suppose the element of is written such that that allocates control to the

lender, then

≡ 1

(A.2)

where is the multivariate normal CDF with mean and variance .

We estimate the strictness using loan covenant ratios in Dealscan and the borrowers’

real financial ratios at the time of issuance from Compusta. We first calculate the slack in the

first period of the contract as the difference between log value of observed ratio and the log

value of minimum covenant ratio (or the negative of the difference in the case of maximum

ratio) specified in the contract. We take 10 commonly used covenants ratios as follows,

minimum EBITDA/debt, current ratio, quick ratio, tangible net worth, total net worth,

EBITDA, fixed charge coverage, and interest coverage, and maximum debt/equity ratio, and

capital expenditure. We then estimate as the covariance matrix associated with quarterly

changes in the log value of financial ratios of US firms which have borrowed from Dealscan

since 1990. To allow for variation from the correlation structure of ratios, we estimate a

separate covariance matric for each one-digit SIC industry.

Several caveats are as follows, first, since the covenants are reported at the package level,

we use the package as the unit of observation when estimating the probability of covenant

violation. Since our measure is based on the 10 commonly used covenant ratios, we keep

only the package observations with at least one recording of 10 covenant ratios. Second,

examining the ratios in log value of financial ratios can extend the support of otherwise

constrained ratios to more closely approximate a multivariate normal distribution for changes

40

in ratios. Third, the proxy of strictness may be subject to measurement error. The covenant

ratio calculation methodology may differ from different contracts. Some may also refer non-

GAAP accounting data presented and certified by the CFO by not available with Compusta or

public. Yet, as argued in Murfin (2012), as long as the contract strictness is treated as

dependent variable, measurement error can be absorbed into the model’s error term.

Moreover, the measurement error is largely driven by borrower-specific factors and therefore

can be subsumed by borrower fixed effects used in the analysis.

41

Appendix II Measuring Excess cash

Following Opler, Pinkowitz, Stulz, and Williamson (1999) and Dittmar and Mahrt-

Smith (2007), we estimate the excess cash of US firms. We first employ the determinants of

cash holdings to determine the normal level of cash holdings. We then define the excess cash

as the difference between actual cash holdings and the fitted value estimated from the

regression model. Specifically, we mainly use the following regression to compute the excess

cash, ,

,,

,

,

,

,,

,

,

,

, ,

(A.3)

Where , is the cash and cash equivalents at time t, , is the net assets which is equal

to the total assets minus the cash and cash equivalents at time t, , is defined as current

assets minus current liabilities minus cash at time t, , is computed as the

standard deviation of cash flow to assets for the previous 10 years, , is the market value

at time t plus the total liabilities, and , is the R&D expenditure, set to zero if missing,

over year t.

Several caveats are as follows, first, as in Dittmar and Mahrt-Smith (2007) and Fresard

and Salva (2010), we examine the value of firm cash holdings using the market to book ratio

as proxy for firm value. Thus, it is potentially biased if we use this measure as an additional

determinant of firm cash holdings. To avoid the endogenous issue, we use the three year

lagged sales growth as an instrument variable for the market to book ratio following Dittmar

and Mahrt-Smith (2007). Second, we add the firm effects in the estimation to solve potential

bias caused by unobservable constant firm factors which may correlate with cash holdings.

We do not deduct the estimated specific firm effects when estimating the excess cash. As

argued in Dittmar and Mahrt-Smith (2007), firm fixed effects do not capture the economic

determinants of firm cash holdings such as investment, hedging, operation needs, and so on.

We employ two stage OLS regressions to calculate the excess cash. We first examine

the effect of sales growth and other control variables on the market to book ratio. We further

42

use the fitted value estimated from the first stage as the instrument proxy for market to book

ratio and estimate the determinants of firm cash holdings using model (A.3). The excess cash

level are calculated as the difference between the actual cash holding level and the fitted level

estimated from the second stage excluding the firm fixed effects. The detailed estimation

results are available on request.

43

Appendix III Additional tests

We investigate the impact of corporate fraud on non-price loan monitoring.

A3 Contract Strictness

We find costly external financing cost evidence from the changes of loan spread around

fraud incident in section 3.1. In this subsection, we aim to further obtain restrictive external

financing evidence by testing the changes of non-price loan items associated with corporate

fraud. In particular, we construct the contract strictness proxy as in Murfin (2012) to capture

the probability of covenant violation. 20 Since corporate fraud results in information

asymmetry and uncertainty issue, we anticipate that banks increase their monitoring in terms

of strict contract ratios.

[Insert Table A.1 here]

In table A.1, we report the effect of corporate fraud on the contract strictness. The

dependent variable is the probability of covenant violation constructed as in Murfin (2012).

See appendix I for details about measure construction. Since the measure is calculated at

package level, we test using package level observations. Following Murfin (2012), we use the

maximum maturity among the facilities as the maturity of the package and use the total loan

amount of all the facilities in one package as the loan size. In column 1, we report the results

by controlling contract features, deal purpose, industry, and year effect. In column 2, we

further add the firm variables into regression. In column 3, we control firm effect. Since the

proxy is a continuous variable between 0 and 1, we further do Tobit regression censoring left

at 0 and right at 1 in column 4. All the estimations are reported robust standard errors

clustered at firm level.

In column 1, we find that the probability of covenant violation increase by 10.7

percentage after corporate fraud. The significance level remains unchanged after we add firm

characteristics in column 2. After we control firm fixed effect in column 3, the effect of

20 Previous studies mainly use the number of covenants to proxy the monitoring role of banks (see Graham, Li, and Qiu, 2008). As discussed in Murfin (2012), the contract strictness measure captures four dimensions of financial covenants including the number, slackness, scale, and the covariance of covenants. Moreover, the measure has a stronger predictive power for covenant violation compared with number of financial covenants.

44

corporate fraud on contract strictness increases to 15.6 percentage. The result is more

significant when we use the Tobit model regression in column 4. The results might indicate

that corporate fraud not only result in costly external finance, but also cause to restrictive

external finance.

For remaining variables, banks tend to increase monitoring for large amount loan

contracts. We also find that firms with high credit risk characterized by small firm size, high

leverage, low profitability, low tangibility, and low investment grade are more likely to face

strict monitoring (the result in column 2). In addition, in the Tobit regression, we also find

that the loan contract strictness is negatively related with credit spread (column 4). To some

degree, high contract strictness might increase the default risk of firms when the economy is

in recessions (high credit spread). Thus, banks may loose their contract strictness to reduce

the default risk of firms.

45

Table A.1 The effect of corporate fraud on contract strictness

This table presents the regression results on the effect of corporate fraud on contract strictness. The dependent

variable is the contract strictness proxy as in Murfin (2012) (package level). The post fraud dummy is defined to

be equal 1 for packages which originate after corporate fraud and 0, otherwise. Definitions of all the other

variables are reported in Table 1. Heteroscedasticity-consistent robust standard errors clustered at firm level are

reported in brackets. Significance at the 10%, 5%, and 1% level is indicated by *, **, and ***, respectively.

(1) (2) (3) (4)

Post fraud 0.107 0.091 0.156 0.156

[0.043]** [0.038]** [0.068]** [0.054]***

Log loan size (package) 0.003 0.053 0.048 0.049

[0.016] [0.016]*** [0.029]* [0.023]**

Log maturity 0.049 0.018 -0.016 -0.016

[0.037] [0.031] [0.044] [0.035]

Price grid -0.006 0.029 0.053 0.053

[0.037] [0.035] [0.051] [0.040]

Revolver -0.104 -0.076 -0.021 -0.022

[0.034]*** [0.034]** [0.043] [0.034]

Num. of lenders -0.002 -0.001 0.001 0.001

[0.002] [0.001] [0.002] [0.001]

Log firm size -0.064 -0.075 -0.076

[0.017]*** [0.060] [0.048]

Leverage 0.363 0.302 0.295

[0.108]*** [0.198] [0.159]*

Profitability -0.331 -0.568 -0.572

[0.200] [0.338]* [0.268]**

Tangibility -0.175 -0.189 -0.181

[0.089]* [0.372] [0.297]

Market to book -0.009 -0.004 -0.003

[0.008] [0.007] [0.006]

S&P ratings 0.039 0.006 0.006

[0.018]** [0.030] [0.024]

GDP growth 0.020 0.033 0.038 -0.051

[0.032] [0.031] [0.041] [0.045]

Credit spread -1.169

[0.382]***

Constant 0.065 -0.608 -0.012 1.382

[0.276] [0.304]** [0.745] [0.853]

Deal purpose Yes Yes Yes Yes

Industry Yes Yes Yes Yes

Year Yes Yes Yes Yes

Firm No No Yes Yes

Observations 476 476 476 476


the financial implications of corporate fraudconference/conference2012/proceedings...this paper...

Documents